Unverified50% confidenceFactExact time
The September version's promotional page said '552B Backbone Parameters', emphasizing backbone parameters; this time the basis was switched from backbone parameters to all weights, and the architecture itself is unchanged.
1
Sources
50%
Confidence
Long-term
Relevance
10/18/2026
First Seen
Sources
DeepSeek V4.1 Flash Update: The Parameter Truth, Benchmark Controversy, and Cost Revolution
bilibiliAI技术投降派10/18/2026
Related Entities
Related Claims
Partially VerifiedStep-5-Preview adopts an MoE architecture with 600B total parameters and 27B activated parameters.61% similarUnverifiedStep 5 Preview's 600B total parameters correspond to about 27B activated parameters, an activation ratio of less than 5%, which is a highly sparse configuration.55% similarUnverifiedUnder the optimal configuration, the EduBehaviors framework achieves a macro F1 score of 0.673 and a Cohen's kappa of 0.688.54% similarUnverifiedFinal results of the four pre-registered hypotheses: H1 (the knob is not the same tool) supported; H2 (more thinking is harmful on distractor counting) not supported; H3 (more thinking helps deductive reasoning) supported only for gpt-5.4-mini; H4 (the bill still rises when accuracy does not) supported (descriptive).53% similarUnverified千问3.8 27B基于千问3.5的整套架构体系，架构层面与3.5完全一致，仅做了继续预训练和后训练（含SFT与RL强化）52% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1059805API
curl https://kongchang.com/api/v1/knowledge/claims/1059805MCP
get_claim(id=1059805)